使用torch.nn.Embedding时出现运行时错误，请求排查

Hey there, let's work through why your Embedding module code is throwing a runtime error. I'll break down the most common issues and fixes step by step:

Common Causes & Fixes #

• Missing required imports
  It's easy to overlook importing the core PyTorch modules. Your code needs these two lines at the top to run without errors:

  import torch
  import torch.nn as nn
  

  Without them, Python won't recognize nn.Embedding or torch.LongTensor, which will immediately trigger a runtime error.

• Index out of bounds
  When you initialize nn.Embedding(10, 3), you're creating 10 embedding vectors, indexed from 0 to 9 (inclusive). Double-check your input tensor—if any value is 10 or higher, PyTorch will throw an index error. Your sample input uses indices up to 9 (which is valid), but make sure your actual code doesn't have any out-of-range values.

• Device mismatch between embedding and input
  If you've moved the embedding module to a GPU (e.g., embedding = embedding.to('cuda')), your input tensor needs to live on the same device too. Add this line before passing the input to the embedding:

  input = input.to(embedding.device)
  

  A device mismatch is one of the most frequent sources of runtime errors with PyTorch modules.

• Outdated PyTorch version
  While less likely, some older PyTorch versions might have subtle behavior differences. You can check your current version with:

  print(torch.__version__)
  

  Updating to a recent stable release could resolve unexpected issues.

Working Example Code #

Here's the full, runnable version of your code with all necessary components included:

import torch
import torch.nn as nn

# Initialize embedding with 10 vectors of size 3
embedding = nn.Embedding(10, 3)
# Batch of 2 samples, each with 4 valid indices
input = torch.LongTensor([[1,2,4,5],[4,3,2,9]])
# Generate embeddings
output = embedding(input)
# Print the result (matches the expected structure from the docs)
print(output)

If you've checked all these points and still hit an error, share the exact error message you're seeing—it'll help narrow down the problem even faster!

内容的提问来源于stack exchange，提问作者Mahsa